Dynamic-preserving qualitative motion description for intelligent vehicles

Abstract

Planning, acting, and recognizing intentions of participants in traffic situations requires the processing of complex spatio-temporal situations. If spatio-temporal information was represented quantitatively it would result in a huge amount of data. We claim that an abstraction to a qualitative description leads to more stable representations as similar situations at the quantitative level are mapped to one qualitative representation. Our approach is evaluated by emulating traffic situations with settings in the Robocup small-sized league.